Page 2 of 2

Re: Samsung Q series TVs

Posted: 06 Oct 2018, 10:45
by RealNC
lossofmercy wrote:https://www.youtube.com/watch?v=oHtXx1JOUpo

Just excited to get this a bit more exposure.
Why are they still having issues with the VRR range? 48-60 is really not good at all for consoles. Why not 0-60 through LFC? If you want it to work well on the XBox X, you really need that, IMO.

Re: Samsung Q series TVs

Posted: 16 Oct 2018, 11:45
by Chief Blur Buster
RealNC wrote:Why are they still having issues with the VRR range? 48-60 is really not good at all for consoles. Why not 0-60 through LFC? If you want it to work well on the XBox X, you really need that, IMO.
LFC requires predictive logic to pre-emptively start early re-refresh scanouts without colliding with the timing of the next frame presentation.

This gets extremely challenging when framerates fluctuate very well. LFC works perfectly if frametimes are consistent, since it can reliably simply pre-emptively trigger LCD duplicate-refresh-cycle scanouts exactly between the real frames (non-duplicate refresh cycles).

However, if a new frame arrives while the monitor is still in the middle of a refresh cycle (whether a new refresh cycle, or simply a re-refreshing scanout), that new frame has to wait. That wait creates a bit of lag & a bit of stutter.

So good LFC logic requires:

1. Very fast scanouts (e.g. 240Hz monitors means re-refreshes are only 1/240sec 4.2ms, and less likely to collide with frame presentation timing)
AND
2. Smart logic to pre-emptively time the re-refreshing pass exactly between previous frame and the predicted time of presentation of next frame.

This gets really challenging with slower-scanning panels (e.g. 60Hz panels which will block for 16.7ms during a repeat-refresh pass) combined with wildly fluctuating frametimes. There are rather severe limitations of doing LFC on both G-SYNC and FreeSync that are predicated by laws of physics.

Engineering Point of View: If you are familiar with anti-collision logic (Ethernet, WiFi in the light of bursty internet traffic) the same collision math problem occurs with LFC in the light of bursty frametimes. It cannot be solved perfectly, just improved -- via (1) and (2) as aggressively as possible. Or you decide that the artifacts of doing true low-Hz native refreshing is the lesser-of-evils than doing LFC (and its frame-vs-refresh collision math problem). That is often finely tuned at ~30Hz or ~48Hz, because flicker artifacts (e.g. low-frequency LCD inversion artifacts) can start to become really bothersome, and actually create visible flickering as LCD decay (pixel fade; loss of the static voltage over time of an unrefreshed pixel) starts becoming visible at lower Hz.

Re: Samsung Q series TVs

Posted: 24 Oct 2018, 01:58
by lossofmercy
To summarize: high max frequency + LRC = good?

Re: Samsung Q series TVs

Posted: 25 Oct 2018, 18:27
by Chief Blur Buster
LFC, not LRC.

LFC = Low Framerate Compensation

Having LFC is always good. It's just not a silver bullet. Native low-Hz can de-stutter better.
(But native low-Hz can create other artifacts when a panel is not designed to refresh very well at low Hz..... That's where LFC can be vastly preferred)

However, diminishing returns occur with a high max Hz, which means panel scanouts are really fast, which produces very little stutter disruptions for single-frame LFC mis-predicts (new frame gets forced to wait while an auto-repeat-refresh-cycle finishes refreshing the panel).

LFC can be better than native low Hz beyond a certain point. A panel can start having visual artifacts refreshing at low Hz (flicker and/or low-frequency inversion artifacts) so using LFC instead can be better and instead accepting stutters from from LFC mis-predicts (the inevitable collisions between timings of auto-repeat-refresh and timings of new refresh cycles). A high max Hz and wide VRR range, does make LFC perform much better.

Say, a monitor with 30-240Hz range. That means at 29 frames per second, a worst-case LFC fail (new frame collision with the predictive repeat-refresh scanout action) is only a 4.2ms penalty (1/240sec scanout) on 34ms (1/29sec frametime) -- that is essentially a worst-case stutter amplitude of only (4.2/34)ths -- barely more than a 10% divergence in eyetracking-versus-screenmotion -- so basically an extremely tiny microstutter completely hidden by already harsh-looking low-framerate stutteriness. So LFC mispredicts on 240Hz monitors are not noticeable in most cases. The repeat-refresh stalls are so brief (4.2) that natural frametime fluctuations (GPU fluctuating framerates) are a far bigger factor than LFC mis-predicts.

So yes.... A very high max Hz (e.g. 240Hz) and a very low min Hz (e.g. 30Hz) will produce very seamless LFC (for all framerates below 30fps) that practically looks like native low Hz, for all effective purposes.

Re: Samsung Q series TVs

Posted: 17 Jan 2019, 03:12
by lossofmercy
Ok, so this means that we can officially play these TVs with Nvidia graphics card. Woot Woot.

Re: Samsung Q series TVs

Posted: 19 Jan 2019, 11:36
by RealNC
lossofmercy wrote:Ok, so this means that we can officially play these TVs with Nvidia graphics card. Woot Woot.
Nope. NVidia doesn't support VRR over HDMI. Only DisplayPort.

Re: Samsung Q series TVs

Posted: 20 Jan 2019, 23:40
by lossofmercy
Don't these TVs have displayport?

Re: Samsung Q series TVs

Posted: 20 Jan 2019, 23:43
by lossofmercy
https://www.newegg.com/Product/Product. ... lsrc=aw.ds

Looks like it doesn't. I hope NVidia supports adaptive sync over HDMI 2.1 in the future. The fact that the 2080 or the 2080TI didn't have it was a pretty big disappointment.